A wheel point cloud detection method for smart garage

Abstract

The invention provides a wheel point cloud detection method for an intelligent garage. The method includes the following steps: collecting a three-dimensional point cloud of a parking area scene by using a two-dimensional single-line laser radar; marking a data set; positioning a position of a wheel by an unstructured point cloud based three-dimensional target detection network model; determininga target point cloud range based on a set space candidate frame and extracting a wheel target point cloud of a region of interest; and representing a wheel target by using a spatial minimum outer bounding box based on predicted six degrees of freedom, wherein the wheel target includes position and attitude information of the wheel. The invention provides a deep learning algorithm for wheel targetdetection for the intelligent garage scene. Compared with a conventional target detection method, the method has the characteristics of high efficiency, strong positioning ability, accurate attitude recognition, strong portability and the like and is initially applied in urban intelligent garages.

Description

technical field [0001] The invention relates to the fields of computer vision and artificial intelligence, in particular to a wheel point cloud detection method for intelligent garages based on two-dimensional single-line laser acquisition, which can accurately locate the position of the wheel and recognize the attitude of the wheel. Background technique [0002] Most automated garages use intelligent robots to move vehicles to designated parking spaces. At present, many parking robots on the market can use the limited land area to a certain extent to obtain doubled parking spaces. But the access method that these garages adopt belongs to passive type basically. For example, the pallet-type robot requires the driver to park the car on the vehicle-carrying platform, and then move the vehicle to the corresponding parking space through the handling robot, but the space on the vehicle-carrying platform is small, requiring the driver to have good parking skills, and the vehicle ...

AI Technical Summary

Problems solved by technology

Literature (Wang Hai. Research on automatic parking environment perception technology based on laser radar. Dalian University of Technology, 2013) uses laser scanner to obtain vehicles, walls, pedestrians and other obstacles in the front range, and analyzes feasible parking spaces, but this The method is based on the traditional algorithm to extract vehicle features, which has poor robustness

A vehicle positioning method, automatic driving control method and related systems disclosed in the patent literature (publication number: CN109596140A, application publication), transmit detection signals to the wheels of a moving vehicle, and use laser radar to receive signals reflected by the wheels The way to obtain the distance and direction of the wheel relative to the lidar, this method also requires the vehicle to move along a fixed lane, and cannot recognize the wheel attitude

Images